Fine, S. (2021). An Empirical Look at Psychometric-Based Credit Scoring
A borrower’s personal character has long been considered an important aspect of credit scoring, but is seldom measured by traditional credit models. Psychometric-based credit scoring offers an interesting solution to this issue by providing character-related credit data that can complement traditional models, and supplement the limited credit histories of underbanked groups. And yet, little is known about the empirical validity of psychometric credit scores. The present study attempts to explore this issue by examining the correlates of a psychometric-based credit scoring tool in three diverse geographies, among independent samples of consumer borrowers. The results of this study show the psychometric solution to be an effective predictor of loan defaults, and incrementally valid to traditional credit scores across the samples. As such, the study provides support for the possible use of such scoring solutions in loan underwriting, for which several possible applications are discussed. For example, increasing approvals and credit limits among marginal declines and thin-file borrowers, and personalizing loan servicing.
Fine, S. (2019). Credit Based on Character: The Promise of Psychometric Credit Scoring in Europe.
Traditional credit scores are based almost entirely on historical financial data, and unfortunately become untenable when historical data are sparse. Such is the case for some 40 million unbanked Europeans, and nearly 2 billion unbanked people around the world, who lack credit histories, and who are ostensibly blocked access to affordable credit as a result. In an attempt to address this issue, new alternative credit scoring solutions have gained traction in recent years. One such alternative credit scoring approach, which is not based on historical data, measures creditworthiness based on a borrower’s personal character. Personal character has long been considered an important component of credit scoring, but is rarely measured directly by traditional credit models. Moreover, little is known about the predictive value that psychometric measures of character can have in credit scoring. This paper attempts to explore this issue by studying a psychometric-based credit scoring tool in applied scenarios. Specifically, we will briefly discuss the development and validation of “Worthy Credit”, a brief psychometric questionnaire designed to assess consumer creditworthiness. In addition, empirical data will be presented from two European lenders, demonstrating the predictive validity of the psychometric scores, above and beyond the lenders’ financial-based scores. Finally, we will present a model for evaluating the economic utility of psychometric scores for increased loan approvals among the underbanked.
Fine, S. (2019). Selected Issues for the Evaluation and Implementation of Psychometric Based Credit Scoring
Alternative credit scoring solutions are being embraced by banks and lenders around the world as a means to improve their credit models. While these solutions present exciting new business opportunities, particularly for servicing the underbanked, some solutions include cross-disciplinary technologies, which apply non-financial data to credit scoring, and are less widely understood among credit risk specialists. Such is often the case with psychometric-based credit scoring, for example, which may show promise as a predictor of loan repayments, but has unique characteristics which are critical for their evaluation and implementation. This paper will briefly highlight a few key topics, based on best practice guidelines in psychometric testing, for the evaluation and implementation of psychometric scoring solutions, while providing examples from field studies. In terms of test evaluations, we will cover: Theory-based measurement models, measurement issues such as test reliability and validity, test development and design, testing accommodations, privacy issues, cultural differences, adverse impact, fakability, and more. In terms of test implementations, we will cover: The initial audit; setting realistic objectives; choosing the right tool; positioning the tool in the application process; piloting; integrating scores and operational usage, and data monitoring.